[bananas] Second and Third patches for 15OCT98
Pat Murphy
pmurphy at nrao.edu
Tue Jan 19 23:25:12 EST 1999
Here are the second and third patches for 15OCT98:
Task Description of problem Date
-------------------------------------------------------------------
Many Gridded Clean Component subtraction in error 1999-01-05
FITLD Init error when appending to >2 Gbyte files 1999-01-19
Complete information on these, including links to the modified source code
and instructions for rebuilding the affected tasks, can be obtained via:
http://www.cv.nrao.edu/aips/15OCT98/patches.html
The text of the equivalent README.15OCT98 patch file is also appended
here.
----------------------------------------------------------------------
ID: 2.
Task: Gridded Clean-component modeling
Summary: A bug has been found in the code which subtracts a Clean
Component model from uv data using the gridded method.
The term which specifies the Z separation of a field
from the phase center (sqrt (1 - xshift**2 - yshift**2))
was supposed to be scaled by the ratio of the frequency
of the first spectral (or IF) channel in the model
subtraction to the reference frequency. This scaling
factor was omitted from the code, apparently forever or
at least since 1990. The bug can be important. In a
test case, with data at declination -26, a strong
interfering source 1 degree north of the phase center,
and a frequency difference of 20%, the Clean in IMAGR
was quickly divergent. Note that this test involves a
large frequency difference, a largish separation, and a
declination and direction of separation that emphasize
the W * Z term.
There was no error when the frequency of the first
channel/IF was the reference frequency. This is usually the
case in multi-IF bandwidth synthesis with BIF = 1. In
spectral-line, single-channel synthesis the frequency
differences are normally very small. And, at more northerly
declinations or field separations with horizontal parts of
the vector, the W * Z term is less important. Perhaps this
is why this error was not reported and found until now.
Files: /15OCT98/Q/SUB/NOTST/ALGSUB.FOR
/15OCT98/Q/SUB/NOTST/ALGSTB.FOR
Fix: 1. Fetch the files listed above and place them in
the corresponding area ($QNOT) on your system.
2. COMRPL $QNOT/ALGSUB $QNOT/ALGSTB
3. COMLNK $QPGOOP/{BLING,GPHAS,IMAGR,SCMAP,WFCLN}
COMLNK $QYPGNOT/MX
COMLNK $QPGNOT/{BLCAL,BPASS,CALIB,CPASS}
COMLNK $QPGNOT/{FRING,KRING,LPCAL,PCAL}
COMLNK $QPGNOT/{UVADC,UVSUB}
Documentor: Eric W Greisen
Date: 1999-01-05
----------------------------------------------------------------------
ID: 3. Appending to very large files
Task: FITLD
Summary: A bug in the routine that initializes uv data I/O was
found when it attempts to append to a file that is already
larger than 2 Gigabytes. This only applies on HPs, DECs,
and those Solaris 2.6+ systems compiled to allow such large
files. UVINIT is used everywhere, but FITLD is the main
place that this problem would be seen.
Files: /15OCT98/APL/SUB/UVINIT.FOR
Fix: 1. Fetch the file listed above and place it in the
corresponding area ($APLSUB) on your system.
2. COMRPL $APLSUB/UVINIT
3. COMLNK $APGNOT/FITLD
Documentor: Eric W. Greisen
Date: 1999-01-19
----------------------------------------------------------------------
--
\|/ ____ \|/ Pat Murphy; email http://orangutan.cv.nrao.edu/kippure.html
@~ / oO \ ~@ Division Head, NRAO/Charlottesville Computing
/_( \__/ )_\ Plan, PGP, Pages: http://orangutan.cv.nrao.edu/plan.html
\_U__/ "The truth points to itself" - Kosh
More information about the Bananas
mailing list